23EEE213: INTRODUCTION TO

MACHINE LEARNING
LECTURE 1

Mr. Nakkeeran M. (Ph.D.)

(Smart Grid Cybersecurity)
1
Faculty Associate
Department of Electrical and Electronics Engineering
Amrita School of Engineering,
Amrita Vishwa Vidyapeetham, Amrita Nagar, Coimbatore
AGENDA

 Machine Learning : What , Why?

 Real world Examples

2
WHAT IS COMMON IN ALL THESE?

3
WHAT IS COMMON IN ALL THESE?

Machine learning replicate activities

4
requiring human cognition
PC: Google Images
MORE EXAMPLES…..

5
 Me
Abi

WHAT IS MACHINE LEARNING?

6 Let’s Learn Machine Learning with Abi and her brother

WHAT IS MACHINE LEARNING?

 Arthur Samuel(1959):
"the field of study that gives computers the ability to learn without being
explicitly programmed."
This is an older, informal definition.

7
WHAT IS MACHINE LEARNING?

Me
Abi

 Abi: Well, that’s the buzz word everyone is talking about. Can you please let me know what
it is all about machine learning?
 Me: Machine Learning is learning from data.
 Abi : Learning from data? I don’t get it. You are talking to a layman so please talk in simpler
terms.
 Me: Imagine you are going to bowling and its your first day, so what do you think how will
you perform?
 Abi : I think very bad, I remember the first time I bowled I threw all the balls in the gutter.
 Me: And what about today?
8
WHAT IS MACHINE LEARNING?

Me
Abi

 Abi: Well, today I think I can easily drop at least 7 in one shot. But what this has to with
machine learning?
 Me: I’ll explain, but how did you improve your game?
 Abi: Well, its simple. I practiced and with time I got experience, I got to know that I need to
adjust the angle at which I am throwing the ball. Also I need to pick a lighter ball for better
accuracy.
 Me: Well, so over time you got experience by tweaking some parameters and your game
improved. Right? This is machine Learning.

9
WHAT IS MACHINE LEARNING?

 Herbert Alexander Simon:

“Learning is any process by which a system improves
performance from experience.”

 “Machine Learning is concerned with computer programs that

automatically improve their performance through experience.”

Learning by Machines
Machines → Computers
Learning → History

10
WHAT IS MACHINE LEARNING?

 Tom Mitchell :
"A computer program is said to learn from experience E with respect to
some class of tasks T and performance measure P, if its performance at
tasks in T, as measured by P, improves with experience E."

Tom Mitchell established the world’s first Machine Learning Department at

Carnegie Mellon University in 2006. He was appointed as the first chair of
the department.

“Machine Learning is concerned with computer programs that

automatically improve their performance through experience.”

11
 WHAT IS MACHINE LEARNING?
 Tom Mitchell :
"A computer program is said to learn from experience E with respect to
some class of tasks T and performance measure P, if its performance at
tasks in T, as measured by P, improves with experience E."

Example: playing checkers.

E = the experience of playing many games of checkers

T = the task of playing checkers.
P = the probability that the program will win the next game.

12
 WHAT IS MACHINE LEARNING?
Suppose your Email program watches which emails you do or do not mark
as spam , and based on that learns how to better filter spam. What is the
task T in this scenario?

A. Classifying emails as spam or not spam.

B. Watching you labels emails as spam or not spam

C. The number(fraction) of emails correctly classified as spam or not spam.

D. None of the above : This is not a machine learning problem

13
 WHAT IS MACHINE LEARNING?
Suppose your Email program watches which emails you do or do not mark
as spam , and based on that learns how to better filter spam. What is the
task T in this scenario?
"A computer program is said to learn from experience E
with respect to some class of tasks T and performance
measure P, if its performance at tasks in T, as measured by
P, improves with experience E."

A. Classifying emails as spam or not spam.→ T

B. Watching you labels emails as spam or not spam →E

C. The number(fraction) of emails correctly classified as spam or not spam.→P

D. None of the above : This is not a machine learning problem

14
WHAT IS MACHINE LEARNING?
 Examples
i) Handwriting recognition learning problem
• Task T : Recognizing and classifying handwritten words within images
• Performance P : Percent of words correctly classified
• Training experience E: A dataset of handwritten words with given classifications

ii) A robot driving learning problem

• Task T : Driving on highways using vision sensors
• Performance measure P : Average distance traveled before an error
• Training experience E: A sequence of images and steering commands recorded
while observing a human driver

iii) A chess learning problem

• Task T : Playing chess
• Performance measure P : Percent of games won against opponents
• Training experience E: Playing practice games against itself
15
WHAT IS MACHINE LEARNING?

Me
Abi

What?
 Abi: Well, Ok! But how do machines experience ? They are not humans!
 Me: Yes, Data :) Experience E corresponds to historical Data. Task T is to hit
the bulls eye in one go and P is the number of pins you drop in one go!
 Abi: Hmm, but what do you mean by learn? Machines don’t have brain as
well.
 Me: Exactly, this is where MATHEMATICS comes into picture! Machines
learn by some complex mathematics concepts, every data for them is in the
form of 0 and 1.
 We don’t want to code the logic for our program instead we want a machine 16
to figure out logic from the data on its own.
 WHAT IS MACHINE LEARNING?

Me
Abi

 Abi: What do you mean that we do not need to code the logic?
 Me: We do not have to write hardcode rules as we write in any
other programming language. For example: Suppose I have this
small data about Payroll of employees in the company:

 Now if you need to find something from this data, what can you
infer?

 Abi: I can observe that when experience is less than equal to 10

years, the person is getting salary x 1.5 times the experience
and this factor increases to 2 when the experience increases
from 10 years.
17
 Me: Correct how will you code this? Say if you need to predict
the salary for a person with 8 years and 16 years of experience?
WHAT IS MACHINE LEARNING?

Me Abi

• Abi: But this was pretty easy, even I could figure out the
relations in the data.
• Me: Well when the scale of the problem increases

if (experience < = 10)

{ salary = experience * 1.5 *
100000}
else if(experience >10) Now, can you find relation between experience,
{ salary = experience * 2 * 100000} job level, rare skill and Salary?
Me: We don’t do this here! We don’t write if,
else in ML as we are not focused on writing
Algorithms.
18
WHAT IS MACHINE LEARNING?

Me Abi

 Abi: It is kind of difficult now!

 Me: This is where machines come into picture! The factor of 1.5 or 2 which you
just calculated in previous example is called weight

Use Machine Learning to find out how

much salary of an employee is
dependent on Experience, Job Level
and Skill
Salary = Experience * Magic_Number_1 +
The columns in yellow are called features
JobLevel * Magic_Number_2 + Skill *
and the column in red is called label.
Magic_Number_3 19
So we calculate weights of features that
contribute in deciding the label.
WHAT IS MACHINE LEARNING?

Me Abi

 Abi: But you only said we do not code the Algorithm and our focus
in on the data.
 Me: Actually Data + Algorithm = Insights. Algorithms are already
developed for us and we need to know which Algorithm to use
when. This was just intuition of one type of Machine Learning, the
field is too big with lots of maths :)

20
 WHY MACHINE LEARNING?
21
WHY MACHINE LEARNING?

Me Abi

 Me: How many hours do you use Facebook daily?

 Abi: about 1 hour daily.
 Me: Do you know Facebook is using Machine Learning to tag your friends in your
uploaded picture? Do you know Facebook shows custom video recommendations
based on your interest history?
 Abi: Ohh! So is this the reason not all news feeds are same?
 Me: Exactly, Facebook knows your likes and dislikes based on historical data and
ML ofcourse :).

22
WHY MACHINE LEARNING?

Me
Me Abi

 Me: Which mailbox you are using?

 Abi: I am using Gmail.
 Me: Do you know how does gmail knows that a particular mail has to go to
priority, social, promotion or spam box ?
 Abi: Obviously your machine learning!
 Me: Yaa, it has millions of emails tagged by the users in a particular category. So
by use of ML they observe the patterns and classify accordingly. Even the keyboard
you are using today uses ML to predict your next word.

23
WHY MACHINE LEARNING?

Me Abi

 Abi: So is it? Are we providing data to Google everytime we type in?

Me: Yep, and google is doing some amazing things with that! One such
example was in Google IO 2018. Have a look.

 Abi: Great! So google incorporated Morse Code into Google Keyboard and it
knows the conversions now due to presence of large translated data. 24
Me: Yup!
WHY MACHINE LEARNING?

Me Abi

Morse code is a method of

communication that uses a
series of short and long
signals, called dots and
dashes, to represent letters,
numbers, and punctuation:

25
WHY MACHINE LEARNING?

Me Abi

 Me: Which website do you use for shopping?

 Abi: Amazon! Its amazing!
Me: Its amazing but most amazing thing is the amazon’s recommendations!
 Abi: So very like we have video recommendations, we have product
recommendations using Machine learning?
Me: Exactly, but the conversion rate here matters a lot. Amazon’s
recommendation engine has fetched it almost a revenue of 35 billion dollars!

26
WHY MACHINE LEARNING?

Me Abi

 Me: So where are we watching the video?

 Abi: YouTube ofcourse! and again are you giving me example of recommendations in
youtube?
Me: No this time a bit difficult! You know how much content in hours is uploaded on
youtube in 1 minute? Take a wild guess
 Abi: Hmmm! 10?
Me: It is close to 500 hours! Do you think YouTube has this much staff so that they can
review this much content for child abuse, violence and copyright?
 Abi: Omg! No not at all!
Me: When scale of problem increases and we have more data, it turns out machines are
only our last hope that is why we need Artificial Intelligence!
 Abi: Great! but all those companies are big 4, my company will never use this technology so
why is this more important now!
Me: Machine Learning is not just a buzzword but an amazing technology that can see 27
such things which are not visible to naked eye.
 WHY NOW?

Source: Syntax Technologies

 Flood of available data (especially with the

advent of the Internet)
 Increasing computational power
 Growing progress in available algorithms and theory developed by researchers 28
 Increasing support from industries
Image: Linked In | Machine Learning vs Deep learning
AI AND ML

 AI is the study of how to train the computers so that computers can do

things which at present human
can do better

 ML is an application of AI that provide system the ability to automatically

learn and improve from experience.

29
COMPONENTS OF MACHINE
LEARNING

30
LEARNING VS INTELLIGENCE

Artificial The name of a whole

intelligence knowledge field, similar to
biology or chemistry.
Machine A part of artificial
Learning intelligence
Neural Are one of machine
"will neural nets replace machine learning" Networks learning types
sounds like "will the wheels replace cars".
Deep A modern method of
Learning building, training, and
using neural networks

31
MAIN DIRECTIONS IN MACHINE
LEARNING

32
CLASSICAL MACHINE LEARNING

33
THE MAP OF THE MACHINE LEARNING WORLD

34
 THANK YOU
35